Georges-Pierre Bonneau

Georges-Pierre.Bonneau at inria.fr
snail mail:
INRIA Rhône-Alpes
655 Av. de l'Europe
F-38330 Saint-Ismier (France)
(+33) 476-615-406 (Voice)
(+33) 476-615-440 (Fax)
Professor at Grenoble Universités.
Research at Lab. LJK and INRIA in the MAVERICK group

red line

[Home] [Publications] [Activities] [Teaching] [Short CV]

red line

MoSIG M2 Project

Title: Exaggerating using vector fields: tools and applications

Supervisors: R. Vergne, G.-P. Bonneau

link to Videos showing the exaggeration with varying magnification parameters for the gradient:

Portrait video

Ostrich video

Input image
"Exaggerated" image using the gradient field

Abstract: Exaggeration is one of the main principles in hand-drawn and computer animation. It is also well known from caricatures. Recently, caricature through exaggeration has been proposed as a tool for the visualization of deformation and the comparative visualization [RGV06]. Inspired from perceptual studies, the student will investigate the exaggeration using vector fields, and explore potential applications in visualization, animation and image and video processing.

Subject: Given a data set and a vector field derived from this data set, our objective is to compute an “exaggeration” of the data based on the vector field. We take inspiration from perceptual studies, showing that the human visual system is likely to use such flow fields to estimate texture [B03] and shape [FTA04] properties. A very simple example consists in deforming an image based on its gradient vector field. The following figures and videos illustrate such an example. The gradient of the original image on the left is computed. The value of each pixel in the final image on the right is taken to be the value of a pixel of the original image at a position translated by the gradient vector field. Magnifying the gradient field accentuates the caricature. The full size images may be seen in the online version of this document. The effect of different vector field will have to be studied. This will be particularly challenging for dynamic data sets, e.g. in animation or video processing. The implementation work will be facilitated by the use of a software library. One issue will be to evaluate bounds for the interval inside which the magnification can be carried out without resulting in artefacts in the exaggerated object. Potential applications will be investigated in three areas. One application is in the field of visualization, where these techniques will be used to give an emphasized visualization of the main features of a data set. In animation, we plan to apply these techniques to exaggerate the movement of deformable objects. In image and video processing, these techniques may be applied as explained earlier using gradient vector fields. Emphasizing the image or video using vector fields should results in possibly interesting visual effects.

[RVG06] P. Rautek, I. Viola, E. Gröller, Caricaturistic Visualization, IEEE Transactions on Visualization and Computer Graphics,Volume 12 Issue 5, September 2006, pp 1085-1092
[B03] O. Ben-Shahar, The perceptual organization of visual flows, Ph.D. Thesis, Yale university, New Haven, CT, December 2003.
[FTA04] Fleming, R. W.; Torralba, A. & Adelson, E. H. (2004), 'Specular reflections and the perception of shape', J. Vis. 4 (9), 798-820.